The Economics Of AI Inference: Electricity Is The New GDP – Yogi Kochhar, Pavan Raju

At the White House last week, Elon Musk and Jensen Huang compressed the economics of artificial intelligence into one line: electricity in, national income out. The scrutiny of the claim is what the article tries to achieve.

"I would bet anyone that a 1 per cent increase in power usage corresponds to roughly a 1 per cent increase in GDP." Elon Musk made that wager on 29 September 2026, seated beside NVIDIA's Jensen Huang at a White House gathering. President Donald Trump had filled the room with executives from Google, Meta, OpenAI and Anthropic.

The arithmetic that followed was brisk.

To ground it: America has roughly 1,200 gigawatts of installed electricity-generation capacity. Its peak demand (the highest load drawn on any single day) runs about 750 gigawatts. But the average load, the electricity consumed at any given moment across a typical year, is approximately 500 gigawatts. Musk's wager pivoted on that average. Five more gigawatts of average consumption would represent a 1 per cent increase, which on a $30 trillion economy works out to roughly $300 billion of annual output, or $60 billion for every gigawatt added.

Huang Supplied The Second Half Of The Argument

The buildings rising across America are no longer data centres, he insisted, because they no longer merely store information. They are factories. Electricity, chips and data go in one end; predictions, designs, code and completed work come out the other. A factory is judged by what it produces, and it produces intelligence that has tangible micro- and macroeconomic impact.

For readers who do not live inside this industry, a dairy farm explains it better than a server rack. Think of raw data as grass. Every hospital record, GST filing, court order, school result, and mandi price is grass: abundant, scattered, and worth almost nothing lying in the field. The AI factory is the cow.

The AI factory consumes data and electricity and produces intelligence. That intelligence becomes economically valuable only when it is embedded in a decision or workflow: a loan sanctioned in hours instead of weeks, a crop advisory reaching a farmer before the rain, or a government file moving without a reminder call.

Read through the farm, Musk's wager says something precise.

A farmer who buys more feed can expect more milk, but only if the cow is healthy and feeds on nutritious grass (read: organised data sets), the milking gets done, and the dairy turns that milk into cheese somebody buys.

Electricity Does Not Become GDP On Its Own

It travels through a chain: reliable power, then compute per watt, then useful intelligence per unit of compute, then adoption inside real workflows, then value someone can measure.

A gigawatt waiting for a grid connection produces nothing. A GPU running a model nobody trusts produces expensive heat. Adoption is the slowest link of all, because it asks institutions to change how they actually work.

This is why the route both men described runs through energy first.

Huang said it plainly: the foundation of the whole stack is power, and America's buildout of 10 to 20 gigawatts a year, with the construction, cooling and pipe fitting it demands, could carry about a million jobs.

Musk went further, pointing to orbital solar arrays feeding compute in space, where the sun never sets.

Both treated power availability as one of the defining constraints of the AI buildout.

Bharat should read that conversation without envy and with a notebook. We hold some of the world's richest grass: population-scale digital rails, multilingual demand, and administrative data flowing from a billion daily transactions.

The IndiaAI Mission has begun building the cow, with access to more than 38,000 GPUs at subsidised rates and roughly 960 megawatts of installed data-centre capacity reported as of February 2026.

Our gap sits at the cheese counter. Health, revenue, courts and agriculture each guard their own silo, so the grass rarely reaches the cow, and too few public workflows are wired to consume the milk.

Sovereign infrastructure can provide the trusted, governed environment in which that scattered data can be brought together with the compute needed to turn it into useful intelligence.

The silver bullet comes from Sunil Gupta@Yotta

The $60 billion figure is a speaker's rounded illustration, not a law of economics, nor can it simply be transferred to India. The direction of the bet is sound, though, and the discipline it demands is ours to adopt: count cheese, not cattle. The country that wins this race will NOT be the one with the most cows. It will be the one whose milk reliably becomes cheese.

Source conversation: https://youtu.be/y4QOc3KzZA0 - USA Today

Yogi Kochhar | President Yotta Foundation | AI Evangelist Pavan Raju | AI Researcher & Technologist

Disclaimer: The views and opinions expressed in this article are those of the author(s) and do not necessarily reflect the official policy or position of the publisher. While every effort has been made to ensure the accuracy of the information, the publisher is not responsible for any errors or omissions, or for the results obtained from the use of this information.